Seismic gas reservoirs can be found by identifying anomalies called Direct Hydrocarbon Indicators (DHI). However, in most cases, such anomalies are difficult to detect due to the large amount of data that must be analyzed, requiring a significant amount of time and specialized human resources. In this work, it is proposed a method that uses a Deep Transformer Neural Network to detect the probability of the existence of DHIs in seismic images. Deep Learning has solved problems that require a lot of time and human effort in less time and with greater accuracy. The best results obtained an accuracy of 98%, a sensitivity of 86%, and a specificity of 98%.